Skip to content

add xccl high priority stream test #2049

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

add xccl high priority stream test #2049

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 26 additions & 0 deletions test/xpu/distributed/test_c10d_xccl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -417,6 +417,32 @@ def _test_broadcast_coalesced(self, process_group, device, root_rank):
if self.rank != root_rank:
self.assertEqual(tensors, target)

def _test_pass_xccl_options(self, pg_opts):
store = c10d.FileStore(self.file_name, self.world_size)
# Test init_process_group accepts options
dist.init_process_group(
"xccl",
world_size=self.world_size,
rank=self.rank,
store=store,
pg_options=pg_opts,
)

# Test with new_group
pg = c10d.new_group([0, 1], pg_options=pg_opts)
# test the process group works as expected
t = torch.tensor([self.rank + 1] * 10).xpu(self.rank)
pg.allreduce(t).wait()
expected_tensor = torch.tensor([3] * 10).xpu(self.rank)
Copy link
Preview

Copilot AI Sep 16, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The magic number 3 in the expected tensor should be calculated explicitly. For a 2-GPU setup (world_size=2), the sum would be (0+1) + (1+1) = 3, but this calculation isn't obvious. Consider using sum(range(self.world_size)) + self.world_size or adding a comment explaining the calculation.

Suggested change
expected_tensor = torch.tensor([3] * 10).xpu(self.rank)
# The expected value is the sum of (rank + 1) for all ranks in the group
expected_value = sum(rank + 1 for rank in range(self.world_size))
expected_tensor = torch.tensor([expected_value] * 10).xpu(self.rank)

Copilot uses AI. Check for mistakes.

self.assertEqual(expected_tensor, t)

@requires_xccl()
@skip_if_lt_x_gpu(2)
def test_pass_xccl_options_high_priority_stream(self):
pg_opts = c10d.ProcessGroupXCCL.Options()
pg_opts.is_high_priority_stream = True
self._test_pass_xccl_options(pg_opts)

@requires_xccl()
@skip_if_lt_x_gpu(2)
def test_broadcast_coalesced_xccl(self):
Expand Down
Loading